The present invention relates to memory devices, particularly dynamic random access memory devices, and, more particularly, to a method and apparatus for preventing defective columns of memory cells from rendering the entire memory device defective.
A conventional memory device is illustrated in FIG. 1 The memory device is a synchronous dynamic random access memory (xe2x80x9cSDRAMxe2x80x9d) 10 that includes an address register 12 adapted to receive row addresses and column addresses through an address bus 14. The address bus 14 is generally coupled to a memory controller (not shown in FIG. 1). Typically, a row address is initially received by the address register 12 and applied to a row address multiplexer 18. The row address multiplexer 18 couples the row address to a number of components associated with either of two memory bank arrays 20 and 22 depending upon the state of a bank address bit forming part of the row address. The arrays 20 and 22 are comprised of memory cells arranged in rows and columns. Associated with each of the arrays 20 and 22 is a respective row address latch 26, which stores the row address, and a row decoder 28, which applies various signals to its respective array 20 or 22 as a function of the stored row address. The row address multiplexer 18 also couples row addresses to the row address latches 26 for the purpose of refreshing the memory cells in the arrays 20 and 22. The row addresses are generated for refresh purposes by a refresh counter 30 that is controlled by a refresh controller 32.
After the row address has been applied to the address register 12 and stored in one of the row address latches 26, a column address is applied to the address register 12. The address register 12 couples the column address to a column address latch 40. Depending on the operating mode of the SDRAM 10, the column address is either coupled through a burst counter 42 to a column address buffer 44, or to the burst counter 42, which applies a sequence of column addresses to the column address buffer 44 starting at the column address output by the address register 12. In either case, the column address buffer 44 applies a column address to a column decoder 48, which applies various column signals to respective sense amplifiers in associated column circuits 50 for the arrays 20 and 22.
Data to be read from one of the arrays 20 or 22 are coupled from the arrays 20 or 22, respectively, to a data bus 58 through the column circuit 50, and a read data path that includes a data output register 56. Data to be written to one of the arrays 20 or 22 are coupled from the data bus 58 through a write data path, including a data input register 60, to one of the column circuits 50 where they are transferred to one of the arrays 20 or 22, respectively. A mask register 64 may be used to selectively alter the flow of data into and out of the column circuits 50 by, for example, selectively masking data to be read from the arrays 20 and 22.
The above-described operation of the SDRAM 10 is controlled by a command decoder 68 responsive to high level command signals received on a control bus 70. These high level command signals, which are typically generated by the memory controller, are a clock enable signal CKE*, a clock signal CLK, a chip select signal CS*, a write enable signal WE*, a row address strobe signal RAS*, and a column address strobe signal CAS*, where the xe2x80x9c*xe2x80x9d designates the signal as active low. The command decoder 68 generates a sequence of command signals responsive to the high level command signals to carry out a function (e.g., a read or a write) designated by each of the high level command signals. These command signals, and the manner in which they accomplish their respective functions, are conventional. Therefore, in the interest of brevity, a further explanation of these control signals will be omitted.
A portion of the column circuits 50 of FIG. 1 is shown in greater detail in FIG. 2. The column circuit 50 is shown connected to a pair of arrays 100, 102, which may be subarrays in either of the arrays 20, 22 shown in FIG. 1. Alternately, a single column circuit 50 containing the circuitry shown in FIG. 2 may be used to access both of the arrays 20, 22 shown in FIG. 1. The column circuit 50 includes a plurality of column node circuits 110a-n in addition to a redundant column node circuit 112. All of these column node circuits 110, 112 are identical, and, in the interest of clarity and brevity, the internal components of only one column node circuit 110a is shown in FIG. 2.
The column node circuit 110a interfaces with two columns of memory cells using two pairs of complementary digit lines D0, D0* and D1, D1*, respectively. However, it will be understood that the column node circuit 110a may contain fewer or greater numbers of complimentary digit line pairs. In the interest of brevity, the digit lines D0, D0* and D1, D1* in the column node circuit 110 as well as in the other column node circuits 110b-n, 112 will sometimes be referred to as simply D and D*. Each digit line pair D, D* has coupled therebetween a negative sense amplifier 120, a positive sense amplifier 122, an equilibration circuit 124, and an I/O circuit 126.
The equilibration circuit 124 is controlled by a precharge control circuit 130 that may be part of the row decoders 28 (FIG. 1) to couple the digit lines D, D* to each other and to an equilibration voltage, which typically has a magnitude equal to one-half the magnitude of a supply voltage. The negative sense amplifier 120 and the positive sense amplifier 122 normally receive respective power signals, typically ground potential and either the supply voltage or a pumped voltage having a magnitude that is slightly greater than the magnitude of the supply voltage, respectively. After the digit lines D, D* have been equilibrated by the equilibration circuit 124, the sense amplifiers 120, 122 detect a voltage imbalance in the digit lines D, D* during a read access of memory cells in the arrays 100, 102. The sense amplifiers 120, 122 then drive the digit lines D, D* in the direction of the imbalance until one of the digit lines is at the supply voltage and the other of the digit lines is at ground potential.
Once the sense amplifiers 120,122 have driven the digit lines D, D* to voltages indicative of the data read from a memory cell in the respective column, the digit lines D0, D1* are coupled to respective I/O lines I/OA, I/OA* by the I/O circuit 126. As is a well understood in the art, in a read memory access the signals from the digit lines are coupled to a DC sense amplifier (not shown), which applies a corresponding data signal to the data bus of the memory device. The other digit lines D1, D1* in the column node circuit 110a are similarly coupled to a respective pair of I/O lines I/OB, I/OB* by a respective I/O circuit 126.
In a write memory access, the I/O lines are driven by respective write drivers (not shown), and are coupled to the digit lines D, D* by the I/O circuit 126.
The column node circuit 110a receives a SEL_R signal from a respective inverter 114 to cause it to couple its digit lines D, D* to the I/O lines I/O, I/O*, respectively. Similarly, the column node circuit 110b receives a SEL_R+1 signal to couple its digit lines to the same I/O lines, and the column node circuit 110n receives a SEL_R+N signal to couple its digit lines to the same I/O lines. Since the SEL signals select various columns of memory cells in the arrays 100, 102, they are normally generated by the column decoder 48 (FIG. 1).
The I/O circuits 126 in the redundant column node circuit 112 are likewise coupled to the same I/O lines by a select SEL_RED signal, but the SEL_RED signal is generated by a redundant column control circuit 144. The redundant column control circuit 144 may be part of the column decoder 48 (FIG. 1).
As mentioned above, the column node circuits 110a-n, 112 are coupled to both arrays 110, 102. However, the column node circuits cannot receive signals indicative of read data from both arrays 100, 102 at the same time. For this reason, isolation transistors 150, 152 are coupled between each digit line D, D* of the column node circuit and corresponding digit lines D, D*, respectively, of the arrays 100, 102. All of the isolation transistors 150 coupled to the array 100 are turned ON by a common ISO_LEFT signal, and all of the isolation transistors 152 coupled to the array 102 are turned ON by a common ISO_RIGHT signal. Since the arrays 100, 102 contain rows of memory cells corresponding to different row addresses, the ISO_LEFT and ISO-RIGHT signals are typically generated by the row decoders 28 (FIG. 1).
Although the manufacturing yield of memory devices is very good, the large number of transistors, signal paths, and other components, such as capacitors, contained in memory devices creates a significant statistical probability that a memory device will contain at least one defective transistor, signal path or other component. For this reason, memory devices typically incorporate rows and columns of redundant memory cells. If a row or column of memory cells is found to be defective during testing, either before or after packaging the memory device, the memory device can be programmed to substitute a redundant row of memory cells for the defective row, or a redundant column of memory cells for the defective column. The redundant column node circuit 112 is provided to interface with redundant columns of memory cells in the arrays 100, 102. The redundant column node circuit 112 interfaces with two columns of memory cells, so that two redundant columns are substituted whenever a single defective column is found during testing. However, it will be understood that redundant columns can be substituted on a column-by-column basis, or that redundant columns can be substituted in groups larger than two. The number of digit lines D, D* in the redundant column node circuit 112 can be adjusted as desired to match the number of redundant columns that are substituted.
Redundant columns of memory cells markedly improve the manufacturing yield of memory devices. However, there are some defects that can occur that cannot be repaired by substituting a redundant column. For example, with reference to FIG. 3, a portion of the arrays 100, 102 includes access transistors 160 coupled between respective digit lines D, D* and a respective storage capacitor 162. Each access transistor 160 selectively couples a digit line D or D* to one plate of the storage capacitor 162. The other plate of the storage capacitor is a xe2x80x9ccell platexe2x80x9d that is typically coupled to a voltage having a magnitude of one-half of the supply voltage. In operation, the storage capacitors 162 store voltages indicative of either a logic xe2x80x9c0xe2x80x9d or a logic xe2x80x9c1xe2x80x9d.
The cell plate of each capacitor 162 is typically common to all of the storage capacitors 162. As a result of manufacturing defects, one of the digit lines D or D* may be shorted to the cell plate either directly (the usual failure mode) or through a shorted storage capacitor 162. During testing of the memory device, this defect will be detected, and a redundant column of memory cells will be substituted for the defective column. However, the sense amplifiers 120, 122 in the column node circuit 110 for the defective column normally continue to receive the NLAT and PSENSE signals from the row decoder 28. The sense amplifiers 120, 122 can thus couple the cell plate to either the supply voltage or ground potential thereby rendering the remainder of the memory cells defective.
Although this problem has been recognized in the past, none of the approaches that have been developed to deal with this problem are entirely satisfactory. One approach has been to selectively decouple the NLAT and PSENSE signals from the column node circuit 110 for the defective column of memory cells. Although this approach does prevent a shorted storage capacitor from rendering the remaining cells defective, it does so at great expense. The transistors that are used to selectively couple the NLAT and PSENSE signals to the column node circuits 110 must be physically very large to provide a sufficiently low impedance path to drive the sense amplifiers 120, 122 so that they can respond with sufficient speed. Driving the sense amplifiers 120, 122 through a relatively high impedance markedly slows the ability of the sense amplifiers 120, 122 to sense voltages on the digit lines D, D*, thereby reducing the access time of the memory device. The amount of surface area on a semiconductor die consumed by adding a relatively large transistor to each negative sense amplifier 120 and a relatively large transistor to each positive sense amplifier 122 is significant because of the large number of the sense amplifiers 120, 122 in a typical memory device.
Another problem with providing transistors to selectively couple the sense amplifiers 120, 122 to the row decoder 28 is the difficulty of routing signal lines in the memory device. More particularly, it would be necessary to supply each column node circuit 110 with two additional signal lines coupled to the gates of the transistors. However, it would be difficult to route this many signal lines to the column node circuits 110.
Another approach to preventing defective columns of memory cells from affecting other memory cells has been to place a laser fuse between each column node circuit 110 and the digit lines D, D* of the arrays 100, 102 to which they are connected. When a column of memory cells is found to be defective during testing, a redundant column of memory cells is substituted for the detective column, and the laser fuse coupling of the defective column to its column node circuit 110 is severed. While this approach has been satisfactory in the past, it is becoming less so because the minimum laser pitch has not kept up with decreases in digit line pitch. Furthermore, while this approach has been satisfactory for repairing defects found before the memory device has been packaged, it cannot be used for repairing post-packaging defects.
Although these problems have been explained with reference to the SDRAM 10 shown in FIG. 1, it will be understood that the same problems exist with other dynamic random access memories (xe2x80x9cDRAMsxe2x80x9d) including asynchronous DRAMs and packetized DRAMs, such as synchronous link DRAMs (xe2x80x9cSLDRAMsxe2x80x9d) and RAMBUS DRAMs (xe2x80x9cRDRAMsxe2x80x9d).
There is therefore a need for a method and apparatus that can be used to repair post-packaging defects in a manner that prevents defective memory cells in a column from affecting other memory cells and which does not unduly increase the cost of memory devices.
A method and apparatus for repairing defective columns of memory cells in a memory device does so in a manner that prevents the defective memory cells from adversely affecting non-defective memory cells. In accordance with one aspect of the invention, a plurality of column node circuits are provided, each of which includes at least one pair of complimentary digit lines. Each of the column node circuits also includes a sense amplifier, an equilibration circuit, and an input/output circuit, each of which is coupled between a respective pair of the complimentary digit lines of the column node circuit. A first pair of coupling switches selectively couples each pair of complimentary digit lines in each column node circuit to a pair of complimentary digit lines for a respective column in a first array. A second pair of coupling switches may optionally be provided to selectively couple each pair of complimentary digit lines in each column node circuit to a pair of complimentary digit lines for a respective column in a second array. The coupling switches each have a conductive state determined by a respective column node disable signal, which is generated by a redundant column control circuit. The redundant column control circuit generates the column node disable signals so that the first and second coupling switches coupled to the respective column node circuits are non-conductive responsive to a redundant column of memory cells being substituted for the column of memory cells to which the column node circuit is coupled.